Strategic Management Journal

Strat. Mgmt. J., 32: 139–158 (2010)

Published online EarlyView in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.868
Received 12 February 2009; Final revision received 5 May 2010

TESTING MANAGEMENT THEORIES: CRITICAL

REALIST PHILOSOPHY AND RESEARCH METHODS
KENT D. MILLER1 * and ERIC W. K. TSANG2
1
The Eli Broad Graduate School of Management, Michigan State University, East
Lansing, Michigan, U.S.A.
2
School of Management, University of Texas at Dallas, Richardson, Texas, U.S.A.

This study identifies the practical and philosophical difficulties associated with testing strategic
management and organization theories. Working from a critical realist perspective, we affirm
the importance of falsification and verification efforts for progress in theory development. We
advocate a four-step approach for advancing theory testing that prioritizes identifying and testing
for the presence and effects of hypothesized causal mechanisms, rather than solely focusing on
correlational methods to jointly test the set of effects composing a theoretical system. Going
beyond prior critical realist writings, we provide practical guidance for deploying established
research methods to test management theories. Copyright  2010 John Wiley & Sons, Ltd.

INTRODUCTION 2007; Pentland, 1999; Weick, 1989) than to test-

Both practical and philosophical obstacles hinder
management researchers’ efforts to test theories
conclusively. Some challenges to testing manage-
ment theories are inherent to the complex, open,
and changing nature of organizations and their con-
texts (Astley and Van de Ven, 1983; Fabian, 2000).
Other obstacles relate to researchers’ beliefs and
practices. For example, imprecisely stating theo-
ries obstructs empirical testing (Astley and Zam-
muto, 1992; Bacharach, 1989). Verifying or falsi-
fying evidence remains disputable due to differing
beliefs about what constitutes relevant evidence
and how such evidence should be collected and
analyzed. In response to such obstacles, manage-
ment researchers have given greater attention to
building theories (Lewis and Grimes, 1999; Locke,
Keywords: critical realism; mechanisms; philosophy of
science; research methods; theory testing
*Correspondence to: Kent D. Miller, The Eli Broad Graduate
School of Management Michigan State University, East Lansing,
MI 48824-1122, U.S.A. E-mail: millerk@bus.msu.edu
ing them (Davis and Marquis, 2005; Hambrick,
2007).
Most empirical studies in strategic management
use correlational methods that do not directly test
the explanatory mechanisms proposed by our the-
ories (Bromiley and Johnson, 2005; Tsang, 2006).
As a result, we fail to provide direct evidence ver-
ifying or refuting our theoretical explanations and
do not rule out alternative explanations. In short,
researchers often fail to test theories rigorously.
As empirical research grows in the management
field and methods proliferate, we need to consider
how our efforts contribute to cumulative progress
toward theory falsification and verification. To sup-
port theory testing efforts, the current study pro-
vides both philosophical arguments and practical
research methods.
We present critical realist philosophy of sci-
ence as an alternative to strict falsificationism
(Popper, 1959) that supports the critical evalua-
tion of theories on the basis of empirical data.
By positing a realist ontology and a fallibilist
epistemology, critical realism motivates efforts

Copyright  2010 John Wiley & Sons, Ltd.

140 K. D. Miller and E. W. K. Tsang
toward falsifying theories in order to advance
scientific knowledge. However, falsifications are
never certain and definitive; they may be revised
(Sayer, 1992: 205–206). Critical realist philoso-
phy acknowledges the potential fallibility of all
knowledge claims, and supports modesty regard-
ing verification and falsification alike. The aim of
researchers evaluating theories is to offer ‘judge-
ments of superiority and inferiority’ (Sayer, 1992:
206).
The current study begins by providing back-
ground on the challenges—both practical and
philosophical—that inhibit the testing of man-
agement theories.1 We then offer an overview
of critical realism, highlighting key aspects per-
taining to the evaluation of theories. However,
the problem remains that prior writings come up
short of translating critical realist philosophy of
science into applied research methods. As Dan-
ermark et al. (2002: 150) succinctly summarize,
‘Critical realism is not a method.’ Acknowledg-
ing this claim, the latter portion of our study
addresses the gap between critical realist phi-
losophy and practical methods for theory testing
within empirical management research. We trans-
late critical realist philosophy into four actionable
research steps. Our approach emphasizes identi-
fying causal mechanisms and testing their effects
directly, as steps complementary to testing theoret-
ical systems using multivariate correlational meth-
ods. Rather than advocating novel methods, we
discuss how to deploy various established research
methods within a critical realist approach to theory
testing.
OBSTACLES TO TESTING
MANAGEMENT THEORIES
The major obstacles to testing management the-
ories in a rigorous manner can be categorized
into four domains: the nature of the social phe-
nomena that management researchers investigate,
imprecise and fragmented theorizing, inadequate
research designs, and inevitable reliance upon
untested assumptions. This section discusses each
of these major obstacles to testing management
theories.
1
By ‘management theories,’ we mean strategic management and
organization theories.
Copyright  2010 John Wiley & Sons, Ltd.
Nature of social phenomena
In the social sciences, ‘it is unrealistic to assume
that all relevant data will be consistent with a the-
ory even if the theory is correct’ (Lieberson, 1992:
7). There are several factors that make conclusive
theory testing more difficult in the social sciences
in general, and management theory in particular,
than in the natural sciences.
First, organizations are diverse, complex, and
changing social phenomena, with multiple levels
of analysis, as well as multiple and contingent
causal processes (Astley and Van de Ven, 1983;
Fabian, 2000). Owing to our inability to identify
all the pertinent variables, we are often unable
to state precisely the conditions on which differ-
ent types of behavior depend or even the contin-
gencies that make particular theories relevant to
certain empirical contexts. Where boundary con-
ditions are ambiguous, it may not be possible
to determine whether theoretical explanations are
direct competitors. From a theory testing view,
the availability of competing hypotheses derived
from different theories is a fortuitous situation,
but it occurs infrequently in management research
due to the diverse range of phenomena studied.
By focusing on diverse phenomena, researchers
tend to position their theories as complementary
rather than competing. By contrast, the history
of natural science shows that testing compet-
ing hypotheses is an effective way to determine
the relative merits of alternative theories (Losee,
2005).
A second factor that renders theory testing dif-
ficult concerns the element of personal volition
in human behavior. Though constrained by habits,
rules, routines, and institutions, humans have free-
dom to choose their actions (Archer, 2000; Down-
ward, Finch, and Ramsay, 2002; Searle, 2001). In
strategic management, this belief is reflected in
arguments supporting strategic choice as causal,
and not merely epiphenomenal (Child, 1997). A
similar conviction is found in research on cre-
ativity in entrepreneurial venturing (Sarasvathy,
2001). Although contextual factors have important
influences on organizations, Child (1972) submits
that organizational decision makers are not pas-
sive. They exercise choice and take actions that
enact their organizations’ environments. Not only
are strategic behaviors nondeterministic, they can
be intentionally unpredictable and deceptive in
efforts to gain advantages over rivals. Freedom and
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

creativity curtail the predictive power of theories
of human behavior.
Third, researchers’ activities may change the
beliefs and practices of managers and thereby
undermine the continuity of the phenomena inves-
tigated (Numagami, 1998). Self-fulfilling and self-
defeating prophecies are not causes for concern
in the natural sciences. A theory of, say, plan-
etary motion will not change how planets actu-
ally move. By contrast, managers may alter their
behaviors on the basis of the knowledge cre-
ated by researchers (Knights, 1992). The extent
to which research influences managerial practice
likely varies widely depending on the theory and
organization of interest.
Moreover, researchers and managers are related
to one another such that researchers themselves
can, at times, be direct causal agents in organiza-
tional actions (Bradbury and Lichtenstein, 2000).
For example, making explicit the implicit theory-
in-use within organizations can catalyze organiza-
tional change (Argyris and Schön, 1978), thereby
undermining the behavioral relevance of a pre-
viously identified theoretical explanation. Con-
versely, an invalid theory, if vigorously promoted
to managers and given sufficient time, could be-
come a more accurate description of their behav-
ior (Brennan, 1994). Ferraro, Pfeffer, and Sutton
(2005, 2009), for example, maintain that economic
theories become self-fulfilling as their behavioral
assumptions become normative in organizations.
A related issue is that unlike the natural sciences,
falsification of a theory in the social sciences
may not eliminate it permanently because the his-
torical and institutional conditions that originally
supported the theory, though subsequently super-
seded or transformed, may, with further historical
change, return again to restore support (Hutchison,
1988).
Theorizing
One of the primary obstacles to testing man-
agement theories is their imprecision. Bacharach
(1989: 501) comments that management theories
‘are often stated in such a vague way that the the-
orists can rebut any discrediting evidence.’ Astley
and Zammuto (1992: 446) note: ‘while linguistic
ambiguity increases the potential number of empir-
ical tests conducted on a theory, it also reduces the
chance that those tests can amount to a refutation
of the theory.’ Imprecise wording of hypotheses
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 141
can make them logically nonfalsifiable. For exam-
ple, Perrow’s (1994: 216) contention that ‘no mat-
ter how hard we might try, the characteristics of
complexly interactive and tightly coupled systems
will cause a major failure, eventually’ cannot be
falsified due to its open time horizon. Imprecise
wording is not unique to strategic management and
organization research. Kuhn (1962) and Lakatos
(1970) documented scientists’ tendency to state
and interpret theories in ways that make them
immune to falsification.
Donaldson (1995) portrays management theories
as oriented toward different levels of analysis and
different value assessments, as well as having dis-
tinct languages and methodologies. For each the-
ory, there is a dedicated group of scholars working
on research to test hypotheses derived from their
theory. To establish their distinct niches within
a research field, groups of researchers empha-
size how their theories differ from one another
(Mone and McKinley, 1993). Over time, barri-
ers between groups of theorists grow (Aldrich,
1992). As researchers perceive their theories to be
incommensurable and discontinue conversations
with advocates of alternative theories (Mahoney,
1993), integration becomes very difficult (Jack-
son and Carter, 1991). For instance, in a debate
with finance researchers Amihud and Lev (1999)
and Denis, Denis, and Sarin (1999) concern-
ing the relation between equity ownership struc-
ture and corporate diversification strategies, Lane,
Cannella, and Lubatkin (1999) claim that agency
theory and strategic management present two dif-
ferent worldviews. Such claims of incommensura-
bility imply difficulty in comparing—logically and
empirically—the merits of distinct theories.
Not only do we have a plurality of theories
within management research, there is also no
consensus about the criteria for evaluating theo-
ries. Critical realism (Fleetwood, 2005; Tsang and
Kwan, 1999), positivism (Donaldson, 1996; Wicks
and Freeman, 1998), constructivism (Mir and Wat-
son, 2000), interpretivism (Lee, 1991), and prag-
matism (Powell, 2002, 2003; Wicks and Freeman,
1998) are among the most prominent philosophies
of science reflected in management studies.
In the natural sciences, a common view is
that there can be only one true theory explain-
ing any particular phenomenon. Thus, researchers
espousing competing theories are keen to pro-
vide not only evidence that supports their theories
but also evidence that challenges rival theories
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
142 K. D. Miller and E. W. K. Tsang
(Chalmers, 1999). By contrast, among manage-
ment theorists, it is generally accepted that the
same phenomenon can be explained by different
theories (Allison, 1971; Ghoshal, 2005). If expla-
nations are not mutually exclusive, greater space
is opened up for researchers to create original
theories that provide novel explanations. Taking
the view that various theories can coexist in har-
mony, some researchers advocate examining orga-
nizational phenomena through multiple theoretical
lenses (Nambisan, 2002; Rajagopalan and Spre-
itzer, 1996). For example, in their study of the
differential impact of foreign institutional and for-
eign corporate shareholders on the performance
of emerging market firms, Douma, George, and
Kabir (2006) use agency, resource-based, and insti-
tutional theories, claiming that a multitheoretic
approach provides a more holistic perspective.
Viewed positively, this approach is compelling
given the complexity of organizations and the
lack of a general theory to account for the rich
variety of organizational phenomena. Viewed neg-
atively, it could evidence a failure to carefully
specify and scrutinize the assumptions, explana-
tions, and implications of different theories to
determine whether they are compatible or incom-
patible. Whereas management researchers are not
entirely averse to engaging one another in debates
(Fabian, 2000), these exchanges tend to remain
at a theoretical, and sometimes a philosophical,
level (De Cock and Jeanes, 2006). For instance,
the debate between Ghoshal and Moran (1996) and
Williamson (1996) concerning how and how far
organizations are able to reduce opportunism has
remained at the theoretical level, stimulating few
follow-up empirical studies. By avoiding clearly
stating competing hypotheses and digging into the
evidence, empirical claims within our theories are
never at risk in such debates.
Research design
Because management theories often do not pro-
vide guidance that is precise enough to know if
our models are properly and fully specified, we
cannot assess the nature and extent of misspecifi-
cation. For example, our theories specify causes
and effects, but rarely tell us the precise tim-
ing (Mitchell and James, 2001). A theory may
alert us to causality running in one direction,
but neglect direct or indirect causal relations run-
ning the other way. Rarely are our tested models
Copyright  2010 John Wiley & Sons, Ltd.
derived from mathematical identities, which pro-
vide some assurance against specification errors.
Instead, researchers are left to their own discre-
tion in choosing which variables to include and the
functional forms of their models. Because our the-
ories are incomplete, we augment our models with
control variables to reflect supplemental explana-
tions to the extent that our data provide relevant
proxies. Exercising discretion in model specifica-
tion does not present an inherent bias against falsi-
fication, but it can be an occasion for opportunistic
behavior by researchers who are intent on verifi-
cation. Researchers seeking to verify their theories
can search for specifications that produce theory-
supporting results (Caudill and Holcombe, 1987,
1999; Leamer, 1978).
Most tests of management theories are not con-
ducted under the conditions of a closed system,
which is defined as ‘one in which a constant
conjunction of events obtains; i.e., in which an
event of type a is invariably accompanied by
an event of type b’ (Bhaskar, 2008: 70). Reg-
ularity of events occurs when a phenomenon is
buffered from outside influences, but conditions
of closure are rarely achievable in the nonex-
perimental social sciences. Macro-organizational
phenomena are often not amenable to laboratory
research, although it may be possible to extrap-
olate, mutatis mutandis, from laboratory research
on individuals and groups to the organization level.
Efforts to isolate social phenomena in laboratory
experiments can introduce artificiality, which alters
behavioral responses (Harré and Second, 1972). It
is impossible to specify all the necessary initial
conditions in a test situation, even in controlled
experiments (Caldwell, 1984). The openness of the
social world entails that a causal mechanism that is
under study will not alone determine the course of
events observed by researchers (Peacock, 2000).
We can learn more from disconfirming cases
than from confirming cases, yet our reasoning
and sampling tend to have a confirmatory bias
(Wason and Johnson-Laird, 1972). The originators
of a theory may hold it with such conviction that
they—consciously or unconsciously—pay atten-
tion primarily to supporting results. Observing
this more than a century ago, Chamberlin (1965
[1890]) advocated the method of multiple working
hypotheses: instead of basing empirical tests on
a single theory, researchers should try to develop
tenable hypotheses for the phenomenon of interest
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

from as many conceivable perspectives as possi-
ble. By doing so, researchers are less likely to be
biased by theory when collecting observations and
conducting analyses.
Researchers tend to adopt what Klayman and Ha
(1987: 211) call a ‘positive test strategy,’ that is,
examining instances in which a theory is expected
to hold. ‘A theory-confirming researcher perse-
veres by modifying procedures until prediction-
supporting results are obtained’ (Greenwald et al.,
1986: 220). Journal review processes reinforce
this confirmatory orientation by favoring theory-
supporting results (Feige, 1975; Pfeffer, 2007).
In a similar vein, Aldrich (1992: 36) laments
that management research ‘systematically disdains
the two processes central to scientific endeavor:
attempted replication of previous findings, and
publication of negative findings.’ A positive test
strategy leads to inflated confidence in a the-
ory’s corroborating evidence and generalizability;
it also discourages exploration of possible alterna-
tive explanations.
Unavoidable assumptions
In science and daily life, we operate on the basis of
untested assumptions. No one can be a consistent
skeptic; action requires trust in the veracity of a
guiding theory (Putnam, 1974). Neither theoretical
nor practical knowledge is free of presumptions
(Gadamer, 2002; Rescher, 1988, 2000). For exam-
ple, researchers’ interest in empirical evidence pre-
supposes that our perceptions of data give some
indication of the way things are. Also, our use of
language to communicate our theories and find-
ings assumes that others will interpret our research
in ways that approximate our intended meaning
(Campbell, 1963). As Polanyi (1962) points out,
we cannot possibly doubt all of our beliefs simul-
taneously; we can only doubt some beliefs on the
basis of other beliefs that we hold unquestioningly.
Hence, Polanyi (1962) portrays scientists as oper-
ating from a ‘post-critical’ or fiduciary perspective,
rather than from universal skepticism. Because of
our unexamined assumptions and human fallibility,
and our limited access to data, our conclusions are
always tentative and our generalizations are risky;
nevertheless, researchers venture bold claims about
the relevance of their findings in settings beyond
those studied.
The Duhem-Quine thesis (Duhem, 1954; Quine,
1961) raises a logical concern that threatens to
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 143
undermine the prospects for testing theories. Ac-
cording to this thesis, theories are never tested
alone. Rather, they are tested together with other
untested auxiliary hypotheses, which include back-
ground assumptions and rules of inference. A the-
ory and its associated auxiliary hypotheses together
form a test system. Let H be the principal hypoth-
esis derived from a substantive theory of interest,
A the set of auxiliary hypotheses, and O the obser-
vational consequence entailed by the conjunction
of H and A so that H ∧ A ⇒ O.2 An observa-
tion ‘not O’ implies ‘not (H ∧ A)’ rather than ‘not
H ’ alone. In other words, the whole test system,
instead of the principal hypothesis, is falsified by
the contradictory outcome (Grünbaum, 1960), and
the existence of auxiliary hypotheses confounds
tests of the principal hypothesis. When a finding
does not support a theory, one can argue that the
problem resides in the failure to control for certain
conditions of the theory, rather than in the theory
itself (Nooteboom, 1986).
Consider, for example, the above mentioned
exchange about the relation between equity
ownership structure and corporate diversification
strategies. Finance and strategic management re-
searchers arrive at distinct conclusions concerning
the relation because they employ different test sys-
tems. Researchers in the two fields measure key
constructs differently and use disparate reasoning
to qualify and interpret empirical evidence (Lane
et al., 1999). These differing auxiliary assumptions
lead to the perception that the two research streams
are not mutually informing, and obstruct reaching
a shared conclusion based on the evidence from
past research.
A critical problem posed by the Duhem-Quine
thesis is that when an instance of falsification
occurs, logic alone cannot isolate the elements
of a test system responsible for the falsification.
This problem of underdetermination is particu-
larly serious in the social sciences where there
is little agreement among researchers as to how
certain auxiliary hypotheses should be tested inde-
pendently (Meehl, 1978).
Testing management theories under the condi-
tions of an open, rather than closed, system further
aggravates the problems associated with isolat-
ing the relations of interest from other confound-
ing effects. Specifying the boundary conditions
of a theory is critical to advancing theory testing
2
The logical operator ∧ signifies ‘and.’
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
144 K. D. Miller and E. W. K. Tsang
(Bacharach, 1989). For instance, the Uppsala inter-
nationalization process model proposed by Johan-
son and Vahlne (1977) maintains that firms exhibit
a pattern of incremental commitment to foreign
markets as learning occurs over time. Early state-
ments of the model were vague about its boundary
condition that the model applies primarily to over-
seas expansions motivated by the objective of mar-
ket seeking. If other motives—such as resource
seeking—are dominant, the model does not apply
(Petersen and Pedersen, 1997). For years, lack of
clarity about this boundary condition caused con-
fusion about how far empirical data supported the
model.
CRITICAL REALISM AND THEORY
EVALUATION
For guidance on how to evaluate theories, we turn
to critical realism, which is a growing intellec-
tual movement in the social sciences (Cruickshank,
2003), especially economics (Lawson, 1997), man-
agement (Fleetwood and Ackroyd, 2004), market-
ing (Hunt, 1992), and sociology (Hamlin, 2002).3
This section briefly describes some key elements of
this philosophy of science, including its epistemo-
logical stance, reliance upon explanations stated in
terms of mechanisms, and view that nature consists
of irreducible strata.
Epistemological stance
Critical realism interrelates ontology and episte-
mology. On the one hand, it posits a realist ontol-
ogy, that is, the existence of a world independent
of researchers’ knowledge of it. On the other hand,
critical realism holds to a fallibilist epistemol-
ogy in which researchers’ knowledge of the world
is socially produced. These two claims jointly
3
It is beyond the scope of this article to compare critical realism
with other philosophical perspectives, but we note a few refer-
ences for the interested reader. Sayer (2000: Part II) compared
critical realism and various postmodernist perspectives. Bhaskar
(1998: Chapter 4), a key founder of critical realism, wrote a
critique of positivism and hermeneutics. Manicas (1987: Chap-
ter 12) critiqued positivist empiricism. Danermark et al. (2002:
Chapter 6) argued the merits of critical realism, as a methodolog-
ical guide, over positivism (which is associated with quantitative
methods), as well as hermeneutics and phenomenology (which
are associated with qualitative methods). In strategic manage-
ment, Mir and Watson (2000, 2001), on one side, and Kwan and
Tsang (2001), on the other, debated the merits of constructivism
and critical realism.
Copyright  2010 John Wiley & Sons, Ltd.
motivate the need for and possibility of critically
evaluating theories. As such, they make up the
core of the critical realist philosophy of science.
Because of the existence of an external referent,
knowledge claims may be challenged and their
merits assessed logically and empirically. Bhaskar
(2008 : 43) notes: ‘To be a fallibilist about knowl-
edge, it is necessary to be a realist about things.’
Science is a personal and social process and
nothing in the methods of science guarantees
success in arriving at truth (Miller, 2005). Re-
searchers’ fallibility and unavoidable reliance upon
presumptions rule out claims to indubitable knowl-
edge, but do not preclude knowledge claims in
general (Bernstein, 1983; Polanyi, 1962). Lacking
an indubitable basis for science, we can, neverthe-
less, reasonably assert the veracity or falsity of sci-
entific theories—albeit, not definitively. A critical
realist perspective affirms the possibility of truth-
ful knowing but acknowledges that human limita-
tions undermine claims to indubitable or objective
knowledge.
Critical realism rejects judgmental relativism —
the inability to judge the merits of theories—as a
general implication of researchers’ fallibility and
social situatedness (Sayer, 2000). Critical realism
takes a balanced and modest stance regarding the
prospects for affirming and rejecting theories based
on empirical evidence. As Sayer (2000: 70) sum-
marizes, ‘Just as assessing truth or practical ade-
quacy is not an all-or-nothing affair, neither is
falsification. In practice it is messy, provisional
and often partial, and the empirical observations
are theory-laden—or at least conceptually medi-
ated—though this does not make it impossible for
the theory in question to be contradicted.’
Mechanisms as explanations
Bhaskar (2008) distinguishes between the domains
of the real, actual, and empirical. The real domain
consists of generative mechanisms, which refer to
‘the ways of acting of things’ (Bhaskar, 2008: 14).4
Through enabling or preventing change, mecha-
nisms give rise to events in the actual domain.
Some events are experienced, and some are not.
The empirical domain is made up of events expe-
rienced through direct or indirect observation, such
4
For further background on mechanisms see: Bunge (1997),
Hedström (2005), Hedström and Swedberg (1998), Machamer,
Darden, and Craver (2000), and Pajunen (2008).
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

as detection through instruments. The reality of the
world does not assure our access.5 Our situation,
including our sensory capacities and instruments,
can render aspects of the world imperceptible.
Mechanisms can be—but need not be—stable,
whereas events and empirical outcomes are always
contingent. Only to the extent that mechanisms are
intransient can they serve as explanations across
cases and time.
The implications of mechanisms need not be
actualized or empirically evident. Realized events
are conjunctures of all of the mechanisms operat-
ing in a situation (Bhaskar, 1998). Lawson offers
a similar contention: ‘A demi-regularity, or demi-
reg for short, is precisely a partial event regularity
which prima facie indicates the occasional, but less
than universal, actualization of a mechanism or
tendency, over a definite region of time-space. The
patterning observed will not be strict if counter-
vailing factors sometimes dominate or frequently
co-determine the outcomes in a variable manner’
(Lawson, 1997: 204, emphasis in original). This
observation places great importance on establish-
ing a theory’s boundary conditions, and account-
ing for countervailing mechanisms in order to test
a theory. For example, if an observed choice of
governance structure contradicts the prediction of
transaction cost economics, it is not necessarily
an incident of falsification. Rather, it may indicate
the presence of countervailing mechanisms, such
as tax incentives or government restrictions. There-
fore, testing a theory’s explanatory power involves
going beyond deducing a theory’s implications and
seeking empirical data regarding those implica-
tions; it requires checking whether the proposed
theoretical mechanisms account for the data.
Bhaskar (2008) notes that the openness of social
systems has both extrinsic and intrinsic sources.
Extrinsic closure refers to the standard notion
of controlling for extraneous variables. Intrinsic
closure inhibits individuals from altering their
behavioral responses to circumstances. Recog-
nizing system openness, critical realists (Archer,
1998; Sayer, 1992) generally have dismissed con-
clusive falsification as unattainable within social
science research. Furthermore, Bhaskar (2008:
160–162) cautions against ‘pseudo-falsification,’
which can occur if a theory is not sufficiently
developed to deal with anomalous findings if the
5
The ‘epistemic fallacy’ associates reality only with the experi-
ential, thereby limiting ontology to epistemology.
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 145
data are wrong, or if countervailing mechanisms
cause researchers to conclude prematurely that a
theory is false.
Rather than focusing solely on empirical out-
comes, critical realists seek explanations for con-
tingent relations, understood in terms of causal
mechanisms. As such, critical realism rejects
Hume’s (2000 [1739]) conception of causality
in terms of universal empirical regularities. A
science concerned with potentialities rather than
actualities, proposes and examines real causal
mechanisms. Hence, critical realists seek to test
explanations, not just correlations between observ-
able antecedents and consequences (Carter and
New, 2004). The search for explanations also goes
beyond the pragmatic concern for whether acting
on a theory produces the expected and desired
effect.
Sayer (1992) emphasizes the priority of expla-
nation over prediction in theory testing. The valid-
ity of predictions depends upon ceteris paribus
conditions, which do not apply in open sys-
tems. By contrast, explanations based on causal
mechanisms may function despite their effects
remaining unrealized due to other countervailing
mechanisms. In light of the openness of social
systems, Bhaskar concludes, ‘. . .the rational devel-
opment and replacement of theories in social
science must be explanatory and non-predictive’
(Bhaskar, 1998: 45–46. emphasis in original).
Testing explanations requires examining not only
the causal mechanism of interest, but also the
empirical evidence regarding intervening and coun-
tervailing mechanisms. In this way, claims regard-
ing intervening and countervailing mechanisms
should not be invoked simply as a post hoc means
to bolster a theory in the face of contrary evidence;
they should be tested.
Strata within reality
Critical realism portrays reality as stratified. The
mechanisms operating in each stratum are unique
and, despite being made up of components
(Pajunen, 2008), may exhibit properties that are
not explicable in terms of mechanisms operat-
ing at other levels. It is a conceptual error to
conflate distinct strata (Archer, 1995). Instead,
emergent properties must be treated as depen-
dent upon other levels, but not reducible to the
mechanisms operating at those levels. Social phe-
nomena emerge from biological and psychological
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
146 K. D. Miller and E. W. K. Tsang
strata, but explanations drawn from biology or psy-
chology may be too distant to account for them.
Following such reasoning, Nickel and Rodriguez
(2002) question management researchers’ extrap-
olations from individual-level theories, such as
Kahneman and Tversky’s (1979) prospect theory,
to formulate hypotheses about organizational risk
taking. The appropriate tests for theories involve
examining mechanisms at the level of analysis at
which they occur, thereby reflecting the relevant
emergent properties that would be lost by exam-
ining mechanisms only at lower strata. Machamer
et al. (2000: 13–14) refer to this level of analysis
as where a mechanism ‘bottoms out.’ Mechanisms
should not be reduced to their components if doing
so strips them of essential properties. In adopt-
ing a stratified view of nature, Bhaskar (1998)
rejects reductionist approaches to social science
such as methodological individualist accounts of
social phenomena. Critical realists reject conflation
of levels and reductionism.
Stratification of the world justifies a parallel
stratification of science. Indeed, a stratified view
of reality simplifies our research efforts by allow-
ing us to focus on those mechanisms germane to
the phenomenon of interest, without needing to
examine mechanisms associated with constituent
levels. Such focus should not, however, be taken to
mean that only one level of analysis is required to
explain any given social phenomenon. For exam-
ple, social phenomena often involve both agency,
at the individual level, and structure, at the social
level (Archer, 1995). This multilevel perspective
is evident in research on technological change
(Orlikowski, 1992), strategy (Jarzabkowski, 2008;
Pozzebon, 2004), and entrepreneurship (Sarason,
Dean, and Dillard, 2006) as structuration pro-
cesses. Mechanisms are not always uncovered by
dropping to a lower level of analysis; mechanisms
can operate downward from a higher level or later-
ally within the same level, as well as upward from
a lower level (Anderson et al., 2006).
CRITICAL REALIST METHODS
The issues to which critical realism alerts us tend to
go unacknowledged in the design and implemen-
tation of management research. Most researchers
work from an implicit empiricist ontology, which
conflates the empirical and the real, rather than
a realist ontology informed by Bhaskar’s (2008)
Copyright  2010 John Wiley & Sons, Ltd.
distinction between the real, the actual, and the
empirical. Using variance designs (Mohr, 1982),
researchers often study empirical regularities ex-
pressed as correlations, rather than examining
causal explanations directly. Although research
methods training sensitizes us to the distinction
between closed and open systems, our theories and
empirical tests reflect to only a limited extent the
critical realist emphasis on outcomes as resulting
from contingent conjunctions of mechanisms.
The methodological implications of critical real-
ism for management researchers remain sketchy.
Ackroyd (2004) offers general implications of crit-
ical realism for management research. Tsang and
Kwan (1999) address the role of replicating prior
research in theory development. Tsoukas (1989),
Easton (2000), and Harrison and Easton (2004)
propose critical realist approaches to case study
research. Mingers (2004) critiques researchers’ use
of statistical modeling. Fairclough (2005) explains
a critical realist approach to organizational dis-
course analysis.
Some critical realists (Wuisman, 2005; Yeung,
1997) acknowledge the gap between critical real-
ism’s contribution to philosophy of science and
the dearth of methodological guidance in this lit-
erature. Those who have begun to address this gap
have done so primarily for the purposes of theory
generation and verification, rather than theory fal-
sification. We see a need to augment critical realist
writings by specifying methods for testing theories.
To the extent possible, proposed methods should
address the obstacles to theory testing identified
earlier.
This section advances practical guidance for
evaluating theories—particularly management the-
ories—that would facilitate theory testing from a
critical realist perspective. We first propose a four-
step approach to theory testing, and then advocate
research programs that combine both extensive and
intensive designs.
Testing mechanisms
Within critical realism, to theorize is to propose
mechanisms that explain events. To identify mech-
anisms, researchers rely upon a mode of reason-
ing beyond the two forms traditionally empha-
sized in philosophy of science—deduction and
induction (Wuisman, 2005). Retroduction is the
mode of inferential reasoning that reconstructs the
conditions for the occurrence of an empirical
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

phenomenon (Bhaskar, 2008; Danermark et al.,
2002; Sayer, 1992). Retroduction seeks to identify
the generative mechanisms that generalize beyond
the immediate instance of the phenomenon and
are critical to its occurrence. To abstract from
specific empirical instances to mechanisms, we
attempt to distinguish general and essential con-
ditions that underpin the phenomenon from inci-
dental and nonessential conditions (i.e., spurious
effects).
Less emphasized in the coverage of critical real-
ist methods is the needed complementary process
of testing postulated mechanisms. Sayer (1992)
advocates testing mechanisms directly instead of
testing their observable implications. This view
reflects an analytical shift from empirical corre-
lations to real mechanisms. Whereas correlation
analyses can evaluate the predicted implications
of a theory, they fail to substantiate or refute
a theory’s causal explanation directly or to rule
out alternative explanations. If mechanisms are
to be truly explanatory, they must be amenable
to testing directly (Bromiley and Johnson, 2005;
Tsang, 2006). However, critical realist discussions
identify two characteristics that present formidable
obstacles to testing explanations: (1) mechanisms
may not be directly observable and (2) their effects
are contingent. If critical realists are to engage in
both theory generating and theory testing research,
they must overcome these two obstacles by pro-
viding methods to (1) identify and measure the
relevant mechanisms in play in a situation and
(2) test for conjunctions of mechanisms as expla-
nations for empirical outcomes.
The concern about unobservable mechanisms
has been addressed elsewhere in strategic manage-
ment research. Godfrey and Hill (1995) distinguish
between constructs that are measurement unob-
servable (i.e., there is no instrument appropriately
calibrated) and state unobservable (i.e., observa-
tion perturbs the state). In the former case, the key
emphasis needs to be on advances in measurement.
For the latter case, effort should be directed toward
techniques for unobtrusive data collection (Webb
et al., 2000). The two forms of unobservability are
not unique to critical realist research, and the meth-
ods for addressing both challenges are well estab-
lished, although success in overcoming them is by
no means guaranteed. Mechanisms that are unob-
servable from the researcher’s perspective may be
apparent, nevertheless, to participants within the
research setting. Questioning participants (using
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 147
interviews or questionnaires) can elicit evidence
regarding mechanisms, although self-report mea-
sures are susceptible to various kinds of biases and
must be treated with caution (Schwarz, Groves,
and Schuman, 1998).
Strategic management research tends to move
from theoretical arguments motivating sets of
hypotheses directly to theory testing using multi-
variate models that incorporate proxies for multiple
theoretical and control variables. Such research
seeks to test full theoretical systems using vari-
ance (i.e., correlation) methods. Bolstering this
approach is the concern that regression models
that are less than fully specified produce biased
coefficients. We support the use of multivariate
correlational methods, but believe that they do not,
by themselves, fulfill the exigencies of critical real-
ism. In particular, they neglect the requirement to
attend directly to testing causal mechanisms as part
of the process of verifying or falsifying a theory.
Examining mechanisms directly calls for added
research steps that specify the hypothesized mech-
anisms, test for the presence of these mechanisms,
and determine whether they function as hypothe-
sized. These three steps, together with testing the
full theoretical system, make up our four-step crit-
ical realist approach to theory testing.
Step 1. The initial step involves identifying
the causal mechanisms believed to account for
the hypothesized relations in a study. This step
requires interpretive work at the interface of theory
and the empirical context of interest. Researchers
must resolve ambiguities in their theories and
derive their implications for particular settings.
The goal is a contextualized specification of the
explanatory properties and processes that under-
lie hypothesized causal relations. Because a theory
may propose different mechanisms to explain dif-
ferent phenomena, researchers need to select those
mechanisms that they believe operate in their par-
ticular research setting. More than one mechanism
may be relevant to a given causal relation; fur-
thermore, mechanisms may have complementary
or conflicting implications for a hypothesized rela-
tion. To illustrate the importance of identifying
all relevant mechanisms, Bromiley and Johnson
(2005) point out the varied and conflicting ways
that top management team diversity affects firm
performance.
To detail the steps in our critical realist approach
to testing theories, consider a simple theory con-
tending that variables x1 and x2 jointly determine
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
148 K. D. Miller and E. W. K. Tsang
Step 1: Identify the hypothesized mechanisms.
x1
m1
y
x2 m2
Step 2: Test for the presence of the mechanisms (m1 and m2)
in the empirical setting.
Step 3: Test isolated causal relations.
x1 y
x2 y observable indicators provide indirect support for
Step 4: Test the theoretical system.
x1
b1 y e
x2 b2
Figure 1. Steps in a critical realist approach to testing
theories
y. The top portion of Figure 1 shows the path
diagram representing this theoretical system. The
hypothesized mechanisms, m1 and m2 link x1 and
x2 , respectively, to y. We maintain the impor-
tant distinction between the parameters relating
x1 and x2 to y (designated β1 and β2 in step
4), and the mechanisms themselves, m1 and m2 .
The circle encompassing the path diagram reflects
the explicit or implicit ceteris paribus condition,
which treats the theoretical system as closed. Spec-
ifying a model according to step 1 of Figure 1 chal-
lenges theorists to clarify theoretical mechanisms
(Anderson et al., 2006) and boundary conditions
(Bacharach, 1989).
Step 2. The next step in testing this theory is
to consider whether the proposed mechanisms are
indeed present in the empirical setting. Step 2 of
Figure 1 conveys the emphasis on substantiating
the existence of the causal mechanisms. Failure
to validate the presence or the nature of postu-
lated mechanisms provides compelling evidence to
reject a theory’s arguments. Prioritizing testing for
mechanisms, rather than jumping directly to testing
their hypothesized relations to dependent variables,
runs contrary to the pattern in strategic manage-
ment research (Bromiley and Johnson, 2005).
Copyright  2010 John Wiley & Sons, Ltd.
Tests for the presence of a mechanism can
involve looking for collateral implications of the
mechanism beyond those predicted by the theory
of interest (Goldthorpe, 2001). Even if a particu-
lar mechanism is not directly observable, the more
observable effects that logically are attributable to
the mechanism, the more compelling is the case for
its presence. For example, consider an argument
claiming that experience with an alliance partner
leads to greater willingness to make asset-specific
investments, with growth in trust as the explana-
tion. Trust between alliance partners may not be
directly observable, but it should produce a vari-
ety of manifest effects on negotiations, contracts,
and coordination processes. In this case, multiple
the presence of an unobserved mechanism. The
reasoning here parallels the use of multiple indi-
cators for latent constructs in structural equation
modeling but, in this case, the construct of interest
is a causal mechanism.
Bhaskar (1998) maintains that people’s reasons
can operate as causes, which in turn are respon-
sible for producing or inhibiting change. Such
intentional human behavior provides an opportu-
nity for testing the mechanism of a theory. The
retroduction of intentions relies upon either truthful
disclosure by the individuals involved, generaliza-
tions about similarly-situated others, or projections
of the researcher’s own intentions onto the stud-
ied subjects. In order to infer reasons and evaluate
the subjects’ espoused reasons, we draw heuristi-
cally upon our own self-understanding for insights
into others’ unobservable intentions. Our common
humanity and experiences, rather than a posture
of objectivity, inform the hermeneutical act of
inferring intentions from others’ actions (Gadamer,
2002; Ricoeur, 1981).
Bromiley and Johnson (2005) and Tsang (2006)
contend that behavioral assumptions are funda-
mental to the explanations of management theo-
ries, yet they are often overlooked in empirical
evaluations. As such, it remains unclear whether
the mechanisms really exist and account for the
phenomena as purported. For example, transac-
tion cost economics assumes that opportunism
is a key factor affecting governance costs, and
economizing on transaction costs in turn deter-
mines the governance choice (Wathne and Heide,
2000). This explanation of managerial choice relies
on an implicit view that perceived, rather than
objective, transaction costs are what managers
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

take into account when making their decisions
(Chiles and McMackin, 1996). To test the trans-
action cost economizing assumption, researchers
need to examine how managers actually perceive
and evaluate transaction costs. In other words,
investigating managers’ perceptions and intentions
would provide a direct test of the alleged mecha-
nism. However, in a survey of the research, Tsang
(2006) found that researchers rarely subject the
theory’s core assumptions to empirical tests. This
assumption-omitting way of testing theories fails to
provide direct tests of their explanations. Two stud-
ies by Buckley and Chapman (1997, 1998) asked
managers directly about the internalization and
externalization of corporate activities and found no
evidence of transaction cost economizing, indicat-
ing the absence of the hypothesized mechanism.
Nevertheless, does Buckley and Chapman’s
(1997, 1998) finding really matter as long as gover-
nance structures observed in the empirical domain
are consistent with the assumption of transaction
cost economizing? Yes, it matters if we care to
understand the process giving rise to the outcome.
There is more than one reason for observed gov-
ernance structures to be consistent with expecta-
tions based on transaction cost economizing: both
natural selection and managerial choice are plau-
sible causal mechanisms. An explanation based on
natural selection ‘adopts the population of organi-
zations as the level of analysis, the environment
as the primary selection mechanism that utilizes
some selection criteria (e.g., transaction cost econ-
omizing), the long run as the appropriate time
frame, and an ex post objective view of costs’
(Chiles and McMackin, 1996: 76). In this framing,
firms engage in a random series of configurational
changes, some of which are by accident transaction
cost reducing while others are not. Firms that hap-
pen to arrive at low transaction cost configurations
will succeed relative to those that do not (Buckley
and Chapman, 1997). The surviving governance
structures are generally consistent with the logic of
transaction cost economics regardless of whether
managers’ choices were based on transaction cost
economizing (Robins, 1987). The issue here is that
the natural selection approach invokes a mecha-
nism very different from that of the managerial
choice approach (Hodgson, 1993). As such, iden-
tifying the mechanisms operating in the empirical
setting is an effective way of assessing competing
theories of the same phenomenon.
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 149
Step 3. If the available evidence affirms the pres-
ence of the theorized mechanisms, we move to
testing their causal effects. Prior to attempting to
verify or falsify an entire theoretical system in an
open context, we advocate testing binary or more
complex subsets of relations under controlled cir-
cumstances. For the hypothetical example shown
in Figure 1, we seek tests of the component x1 -y
and x2 -y relations in contexts that isolate each rela-
tion from other effects. Step 3 of Figure 1 depicts
such tests. To do so, we must move from the open
system of the empirical context of ultimate interest
to experimental or quasi-experimental settings.
In spite of the threats to external validity (Guala
and Mittone, 2005), laboratory experiments allow
for testing the effects of mechanisms under con-
ditions similar to those of a closed system and,
due to random assignment of treatments, gen-
erally provide stronger evidence that supports
or rejects hypotheses than do nonexperimental
designs. Many theories invoke mechanisms that
consist of chains of causal links in which one
event leads to another; experiments help to isolate
and test these links. Moreover, experiments enable
researchers to assess the merits of competing the-
ories by, for example, testing competing mediat-
ing relations proposed by those theories (Shadish,
Cook, and Campbell, 2002). Although experiments
have been a popular research method in psychol-
ogy and organizational behavior, they are seldom
used in strategic management research to facilitate
tests of mechanisms (Croson, Anand, and Agar-
wal, 2007). A notable exception is the experiment
conducted by Sutcliffe and Zaheer (1998) to study
the effects of different kinds of uncertainty on
the likelihood of vertical integration. Another is
Schweiger, Sandberg, and Rechner’s (1989) exper-
iment involving managers in dialectical inquiry,
devil’s advocacy, and consensus approaches to
group strategic decision making.
A behavioral simulation involves individuals in
a setting constructed to mimic key aspects of a
naturally occurring organizational situation. The
realism of the constructed setting is the key feature
distinguishing behavioral simulations from labora-
tory experiments. In an interactive behavioral sim-
ulation, researchers control the simulated context
and observe the behaviors and decisions coming
from participants’ interactions with one another.
The method is particularly suited for strategy pro-
cess research (Dutton and Stumpf, 1991). For
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
150 K. D. Miller and E. W. K. Tsang
example, Hough and White (2003) used an interac-
tive behavioral simulation to study the moderating
effect of environmental dynamism on the relation
of decision making comprehensiveness to decision
quality. Gist, Hopper, and Daniels (1998) report
that interactive behavioral simulations remain quite
rare in management research—even among studies
published in organizational behavior journals.
Quasi-experiments share some characteristics of
laboratory experiments, but they occur in field
settings and researchers have less than full con-
trol over the assignment and scheduling of effects
(Shadish et al., 2002). Thus, quasi-experiments
have some merit by offering conditions that ap-
proximate randomly assigned treatments. In
contrast with laboratory experiments, quasi-
experiments trade off some control for exter-
nal validity (Grant and Wall, 2009). Again, with
few exceptions, management researchers outside
the domain of organizational behavior rarely use
quasi-experimental designs. An example of such
exceptions is Szulanski and Jensen’s (2004) inves-
tigation of the role of templates in the replication
of organizational routines, which was based on an
eight-year longitudinal case study that took the
form of a naturally occurring, repeated-treatment
quasi-experiment.
We recommend that strategic management and
organization researchers further explore the poten-
tial of laboratory experiments, behavioral simula-
tions, and quasi-experiments for providing data to
rigorously test the relations invoked in their the-
ories. Because such research designs fall outside
the training typically acquired in strategic manage-
ment doctoral programs, we see opportunities for
collaborating with scholars in fields such as orga-
nizational behavior and psychology where such
designs are in the mainstream.
Step 4. If empirical data corroborate a theory’s
mechanisms and their effects, then the next step
to take in evaluating the theory is to examine
the implications of its mechanisms jointly. This
step moves the analysis from isolated mechanisms
to the entire theoretical system, thereby adding
complexity to the evaluative procedure. At this
stage, we are interested in whether all of the
theory’s mechanisms are necessary and whether
they are jointly sufficient to explain the outcome.
For a complicated management theory, it may not
be feasible to test all of its mechanisms in a single
study. In such cases, a second-best alternative is
to design tests of subsets of mechanisms that are
Copyright  2010 John Wiley & Sons, Ltd.
unbiased by omitted theoretical variables through
choosing research settings where the unmeasured
variables can be assumed to be either invariant or
irrelevant.
The openness of social systems complicates test-
ing jointly a theory’s hypotheses. Although the
theoretical system is closed (see step 1), the empir-
ical contexts of organizations and industries are
not. Step 4 of Figure 1 depicts the open theoreti-
cal system in which outside influences give rise
to unexplained variance that is relegated to the
error term, ε. Here, the error term results from
omitting variables relevant to the empirical context
but outside the scope of the theory itself. Failure
to account for some of the relevant mechanisms
diminishes the proportion of variance explained
by a model and potentially biases the estimated
effects of the theoretical variables. Step 4, in con-
trast with the original theoretical model (step 1),
focuses on estimable partial correlations (β1 and
β2 ) and omits the generative mechanisms (m1 and
m2 ). This switch from mechanisms to correlations
accommodates established multivariate methods.
We acknowledge that fitting empirical data to a
multivariate model treats the process generating
the dependent variable as a black box and, as such,
correlational analysis can only complement other
research that directly tests mechanisms, not substi-
tute for such tests.
Acknowledging potentially relevant outside con-
tingencies, in large sample studies researchers
attempt to include control variables to account
for effects beyond those in their theory and inter-
actions to reflect conjunctions of mechanisms.
However, critical realist reasoning can lead to
an ‘interactionist regress’ (Sayer, 1992: 189) in
which explaining outcomes as contingent conjunc-
tions of mechanisms results in complex interac-
tions that are difficult to understand conceptually
and test empirically. Large-sample analytical meth-
ods require simplifying generalizations regarding
the contingencies affecting observations, whereas
case research can identify and take into account
idiosyncratic contingencies (Nash, 1999). We dis-
cuss next how different research designs contribute
to theory testing.
Research designs
Critical realism is highly pluralist in terms of
empirical research methods. Because different
methods focus on different aspects of reality,
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

combining several methods together in a research
program yields a rich understanding of the phe-
nomenon of interest (Mingers, 2001). Critical
realists distinguish between extensive and inten-
sive designs, and maintain that both approaches
are needed and complementary (Danermark et al.,
2002; Sayer, 1992, 2000). Extensive research tests
empirical generalities across cases and typically
employs quantitative methods. Intensive research
tries to uncover the explanatory mechanisms rel-
evant in particular cases, and is often associated
with qualitative modes of analysis. Mixed designs
combine both extensive and intensive designs,
thereby utilizing their complementary strengths
and weaknesses. Critical realism encourages mixed
designs in efforts to cope with the complex-
ity of studied phenomena, aspects of which may
go undetected by any single research approach
(Mingers, 2006).
Extensive designs. Critical realist concerns about
relying upon extensive designs revolve around
(1) their focus on empirical correlations rather
than specifying explanatory mechanisms, (2) as-
sumptions of model stability and closure, and
(3) reliance upon statistical tests of significance
(Mingers, 2004). Conveying similar concerns,
Sayer observes: ‘Given the disjunction between
mechanisms and events, a strong correlation (or
some other quantitative association) need not imply
causation, nor a weak one absence of a causal or
structural relation. If a theory is to help solve this
type of problem it must postulate causal mecha-
nisms and not merely specify how total variation
in the dependent variable might relate quantita-
tively to variation in the independent variables’
(Sayer, 1992: 194). Even Starbuck’s (2006) prior-
itization of effect size over statistical significance
omits the goal of explanation through identifying
mechanisms and their surrounding contingencies.
Estimated effect sizes capture causal importance
only if models are specified in terms of contin-
gent relations reflecting complex conjunctures of
mechanisms.
However, Ron (2002) offers some counterargu-
ments to address such concerns. He points out
that open systems do not necessarily undermine
empirical regularities and the occurrence of such
regularities can provide insights into the opera-
tive mechanisms. Working from a critical realist
perspective, a researcher uses regression analysis
to demonstrate the effects of theorized causal
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 151
mechanisms after controlling for other mechanisms
that also could affect the outcome. Ron submits
that researchers’ repeated respecification of regres-
sion models exhibits an exploratory search for
mechanisms affecting an empirical outcome. Such
exploratory use of regression analysis seeks to
explain empirical patterns retrospectively, and, for
open systems the result should not be interpreted
as a predictive model, nor do conventional cri-
teria for evaluating the statistical significance of
coefficients for theory-determined models apply.
If our intent is to falsify theories, researchers
ought to pursue model respecifications that under-
mine hypothesized relations, not just post hoc
respecifications that favor the theory (Caudill and
Holcombe, 1987, 1999; Kennedy, 2005; Leamer,
1978) or adjustments in theoretical arguments to
fit findings (Lipton, 1991).
Manicas (1987, 2006) raises the concern that
the causes postulated in critical realist research
are not linearly additive and, thus, violate a key
assumption of regression modeling. Such a critique
misses some possibilities for using multivariate
modeling in critical realist research. For example,
researchers can use regression models with inter-
action terms to test contingent effects. If a hypoth-
esized effect varies over the range of values for a
regressor, then a spline function can be used. Using
regression analysis, relations for which empirical
tests are feasible consist primarily of direct effects
and lower-level (two-way or three-way) curvilinear
effects, rather than more complex contingent rela-
tions. Hierarchical linear modeling allows coef-
ficients to vary contingent upon moderating and
mediating effects at other levels of analysis (Hof-
mann, Griffin, and Gavin, 2000; Zhang, Zyphur,
and Preacher, 2009) and, as such, is another tool
suitable to critical realists’ interest in testing the
effects of intervening mechanisms on theoretical
relations.
For models with complex and dynamic con-
junctions, computer simulation modeling offers a
means to transition from theoretical arguments to
empirical testing. Experimentation with a simula-
tion model generates data for a response surface
reflecting the effects of multiple variables on a
particular outcome of interest. Such data permit
estimation of an algebraic function, often called a
meta-model, which captures the effects of a set of
independent variables on the dependent variable.
Response surface methods estimate a multivariate
equation that approximates the functional relation
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
152 K. D. Miller and E. W. K. Tsang
of explanatory variables to an outcome variable as
generated by a complex simulation model (Law
and Kelton, 2000). Provided relevant empirical
data are available, an estimated meta-model can
be tested. This combination of simulation mod-
eling of complex systems resulting in an empir-
ically testable regression model provides a way
to move from complex (nonlinear) dynamic sys-
tems to empirical tests using established meth-
ods. Such methods can advance empirical testing
in research areas such as organizational struc-
ture and learning, where simulation research has
enhanced our understanding of the relevant mech-
anisms (Carley, 1992; Lin and Hui, 1999; Rah-
mandad, 2008).
Intensive designs. Intensive designs complement
extensive designs by addressing the differences
across cases that one would expect if empirical out-
comes result from conjunctions of multiple mecha-
nisms in open systems. The purpose behind inten-
sive designs is to identify and describe the gen-
erative mechanisms operating in particular cases,
which is often not feasible for extensive designs.
An intensive design emphasizes the collection of
detailed data within one or more cases. The data
are often qualitative, as are the analytic methods.
Established methods for case study, ethnographic,
grounded theory, and action research guide inten-
sive designs.
A distinctive feature of critical realist deploy-
ment of these methods is the emphasis on retro-
duction. Intensive investigation of a case affords a
unique opportunity to evaluate whether the mecha-
nisms proposed by a theory jointly affect empirical
outcomes as expected, while taking into consid-
eration intervening and countervailing contextual
mechanisms. Easton (2000: 212) concludes: ‘Case
research which would wish to lay claim to a real-
ist philosophy should be carried out in a different
way: to be inquisitive, to look for the roots of
things, to disentangle complexities and to concep-
tualise and re-conceptualise, test and retest, to be
both rigorous and creative and above all to seek for
the underlying reality through the thick veil which
hides it.’ For instance, Porter’s (1993) ethno-
graphic study of how racism affected the occupa-
tional relationships between nurses and doctors in
a hospital adopted a critical realist perspective for
identifying mechanisms that generated the racist
events and found that the universalist-achievement
Copyright  2010 John Wiley & Sons, Ltd.
ethos of professionalism was a mechanism coun-
tering the ascriptive nature of racist tendencies.
Intentional sampling is an important step if
intensive designs are to yield evidence verifying
or falsifying theoretical claims. Contrasting cases
can provide evidence from natural experiments
regarding how mechanisms operate under differ-
ent conditions. Intertemporal comparisons through
repeated studies of the same case can reveal how
continuities and discontinuities in mechanisms and
contexts affect outcomes (Harrison and Easton,
2004). Pathological or extreme cases can reveal
conditions where mechanisms are undermined or
unimpeded (Danermark et al., 2002). As Collier
(1994: 165) explains, ‘[B]y seeing how something
goes wrong we find out more about the conditions
of its working properly than we ever would by
observing it working properly.’
Choosing cases that are very different from
one another enables tests that allow researchers
to judge the relative merits of competing theories
(Stinchcombe, 1968). This approach involves find-
ing theories within a particular research area that
make arguments that lead to conflicting empiri-
cal predictions. Setting up direct empirical tests
of conflicting theoretical predictions provides more
compelling evidence regarding their relative mer-
its than does testing the implications of a sin-
gle theory and then invoking the alternative as
a post hoc explanation for unanticipated findings.
Sagan’s (1993) case study of the Cuban missile
crisis is an exemplar of testing competing mecha-
nisms and implications proposed by the two major
theories that explain accidents in hazardous high-
technology organizations—normal accident theory
(Perrow, 1984) and high reliability theory (La
Porte and Consolini, 1991; Roberts, 1989).
Auxiliary hypotheses. The aforementioned
Duhem-Quine thesis informs how we test mech-
anisms and theoretical systems. Because the pres-
ence of auxiliary hypotheses (A) confounds the
testing of any theoretical hypothesis (H ), it is
possible to accept a given theoretical hypothesis
under one set of auxiliary hypotheses and reject the
hypothesis under another set of auxiliary hypothe-
ses. Although the confounding effects implied by
the Duhem-Quine thesis cannot be eliminated,
Sawyer, Beed, and Sankey (1997) suggest two
ways of assessing the effects of auxiliary hypothe-
ses. One method is to conduct sensitivity analyses
where the auxiliary hypotheses are altered. For
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

example, researchers can draw different samples,
use alternative measures, vary model specifica-
tions, or employ dissimilar statistical methods for
fitting and evaluating models. Consistent falsifying
or supporting evidence using alternative auxiliary
hypotheses is more conclusive than if alternative
auxiliary hypotheses produce conflicting findings.
A second method is to test the auxiliary hypothe-
ses directly. Mäki (2000) distinguishes negligibil-
ity assumptions (that effects are small enough to
neglect), joint negligibility assumptions (that a set
of effects is collectively small enough to neglect),
applicability assumptions (identifying contingen-
cies that make effects non-negligible), early-step
assumptions (restrictions that are later relaxed),
and as-if assumptions (involving counterfactual
claims). At the broadest level, Mäki (2000) distin-
guishes between the core and peripheral assump-
tions of a theory: core assumptions indicate the
major causes postulated by a theory while
peripheral assumptions refer to the minor causes.
Although tests of theories center on core assump-
tions, direct tests of peripheral assumptions can
provide further evidence qualifying theories.
Replications that extend prior studies can help
researchers examine the implications of alternative
auxiliary hypotheses by focusing their attention on
one part of the test system at a time. Suppose
the findings of a study did not support the test
system, H ∧ A. Suspecting that a certain auxiliary
hypothesis caused the result, a researcher replicates
the study with an alternative auxiliary hypothesis
(A ) while keeping the rest of the test system
intact. Finding support for the new test system,
H ∧ A , confirms the researcher’s conjecture.6 If
the system is still not supported, the researcher
needs to consider whether H should be rejected. In
a similar vein, Søberg (2005) submits that a series
of experiments can be used to locate the source of
disconfirming evidence.
CONCLUSION
To successfully test theories, management re-
searchers must overcome a formidable set of
6
An assumption here is that the study has to be repeated in order
to examine the auxiliary hypothesis in question; that is, a simple
reanalysis of the prior data will not do. This is the case when,
for example, the auxiliary hypothesis is about the measurement
of a certain variable in a questionnaire survey. In such a case,
a new survey with a different set of measurement items for that
variable is needed.
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 153
practical and philosophical obstacles, including the
complexity and contingency of social phenom-
ena, imprecisely specified theories, the openness of
social systems, and the unavoidability of untested
assumptions. Critical realist arguments regarding
the conditions that make science possible also are
relevant to making theory testing possible in the
face of such challenges (Bhaskar, 2008; Van de
Ven, 2007). Critical realism confronts the complex-
ity of social phenomena by espousing explanations
stated in terms of mechanisms that generalize,
with empirical effects that are contingent. Criti-
cal realist philosophy acknowledges that theories
are social products, but ontological realism pro-
vides some hope of achieving greater precision
over time through testing our theories. To deal
with the openness of social phenomena, manage-
ment researchers can employ experimental and
quasi-experimental designs to complement nonex-
perimental designs. Critical realism also supports
efforts to replicate prior research and directly test
auxiliary hypotheses (Tsang and Kwan, 1999).
We sought to make critical realist philosophy
of science practical by addressing its implications
for testing management theories. We proposed a
series of four complementary steps for testing
strategic management and organization theories:
(1) identify the hypothesized mechanisms, (2) test
for the presence of the mechanisms in the empir-
ical setting, (3) test isolated causal relations using
experimental or quasi-experimental designs, and
(4) test the theoretical system using correlational
methods. Though there have been scattered dis-
cussions of these steps and the methods that they
entail in the literature, we know of no other attempt
to organize them into an integrative research pro-
cess that facilitates theory testing from a critical
realist perspective.
Most empirical studies in strategic manage-
ment, especially those based on extensive research
designs, seek to identify empirical regularities
expressed as correlations, and thus are located at
step 4 of our recommended four-step approach.
Although many research studies also touch upon
step 1 during the development of hypotheses, steps
2 and 3 have been neglected for the most part.
Without examining causal mechanisms directly,
it is difficult to adjudicate conflicting explana-
tions for empirical findings. Even in the face of
cumulative, mostly confirming empirical findings,
omitting tests for the presence and causal effects
of mechanisms leaves theoretical explanations in
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
154 K. D. Miller and E. W. K. Tsang
doubt. Both intensive and extensive designs can
provide such tests.
Identifying and testing causal mechanisms calls
for increased use of designs and methods found
most frequently elsewhere in the social sciences—
such as generating data through laboratory experi-
ments, behavioral simulations, quasi-experiments,
case studies, ethnography, or grounded theory
building in field settings. Cross-disciplinary collab-
orations and drawing guidance from the methods
literature in other academic fields can facilitate
such research designs. Because each of the four
steps requires specific research techniques and no
researcher is well versed in all these techniques,
there is a need for specialization, with different
researchers focusing on different steps, and the
potential for collaboration as researchers coordi-
nate sequential steps or work together on more than
one step.
ACKNOWLEDGEMENTS
We thank Gregory Dess, Joseph Mahoney, Donald
Palmer, Arkadiy Sakhartov, Roy Suddaby, and the
anonymous reviewers for their helpful comments
on earlier versions of this article.
REFERENCES
Ackroyd S. 2004. Methodology for management and
organisation studies: some implications of critical real-
ism. In Critical Realist Applications in Organisation
and Management Studies, Fleetwood S, Ackroyd S
(eds). Routledge: London, U.K.; 137–163.
Aldrich HE. 1992. Incommensurable paradigms? Vital
signs from three perspectives. In Rethinking Organi-
zation: New Directions in Organization Theory and
Analysis, Reed M, Hughes M (eds). Sage: London,
U.K.; 17–45.
Allison GT. 1971. Essence of Decision: Explaining the
Cuban Missile Crisis. Little, Brown and Company:
Boston, MA.
Amihud Y, Lev B. 1999. Does corporate ownership
structure affect its strategy towards diversification?
Strategic Management Journal 20(11): 1063–1069.
Anderson PJJ, Blatt R, Christianson MK, Grant AM,
Marquis C, Neuman EJ, Sonenshein S, Sutcliffe KM.
2006. Understanding mechanisms in organizational
research: reflections from a collective journey. Journal
of Management Inquiry 15: 102–113.
Archer MS. 1995. Realist Social Theory: The Morpho-
genetic Approach. Cambridge University Press: Cam-
bridge, U.K.
Archer M. 1998. Introduction: realism in the social
sciences. In Critical Realism: Essential Readings,
Copyright  2010 John Wiley & Sons, Ltd.
Archer M, Bhaskar R, Collier A, Lawson T, Norrie A
(eds). Routledge: London, U.K.; 189–205.
Archer MS. 2000. Being Human: The Problem of Agency.
Cambridge University Press: Cambridge, U.K.
Argyris C, Schön DA. 1978. Organizational Learning:
A Theory of Action Perspective. Addison-Wesley:
Reading, MA.
Astley WG, Van de Ven AH. 1983. Central perspectives
and debates in organization theory. Administrative
Science Quarterly 28: 245–273.
Astley WG, Zammuto RF. 1992. Organization science,
managers, and language games. Organization Science
3: 443–460.
Bacharach SB. 1989. Organizational theories: some
criteria for evaluation. Academy of Management
Review 14: 496–513.
Bernstein RJ. 1983. Beyond Objectivism and Relativism:
Science, Hermeneutics, and Praxis. University of
Pennsylvania Press: Philadelphia, PA.
Bhaskar R. 1998. The Possibility of Naturalism: A
Philosophical Critique of the Contemporary Human
Sciences (3rd edn). Routledge: London, U.K.
Bhaskar R. 2008. A Realist Theory of Science (2nd edn).
Verso: London, U.K.
Bradbury H, Lichtenstein BMB. 2000. Relationality in
organizational research: exploring the space between.
Organization Science 11: 551–564.
Brennan MJ. 1994. Incentives, rationality, and society.
Journal of Applied Corporate Finance 7(2): 31–39.
Bromiley P, Johnson S. 2005. Mechanisms and empirical
research. In Research Methodology in Strategy and
Management (Vol. 2), Ketchen DJ Jr, Bergh DD
(eds). Elsevier Science: Amsterdam, The Netherlands;
15–29.
Buckley PJ, Chapman M. 1997. The perception and
measurement of transaction costs. Cambridge Journal
of Economics 21: 127–145.
Buckley PJ, Chapman M. 1998. The management of
cooperative strategies in R&D and innovation
programmes. International Journal of the Economics
of Business 5: 369–381.
Bunge M. 1997. Mechanism and explanation. Philosophy
of the Social Sciences 27: 410–465.
Caldwell BJ. 1984. Some problems with falsificationism
in economics. Philosophy of the Social Sciences 14:
489–495.
Campbell K. 1963. One form of scepticism about
induction. Analysis 23(4): 80–83.
Carley K. 1992. Organizational learning and personnel
turnover. Organization Science 3: 20–46.
Carter B, New C. 2004. Introduction: realist social
theory and empirical research. In Making Realism
Work: Realist Social Theory and Empirical Research,
Carter B, New C (eds). Routledge: London, U.K.;
1–20.
Caudill SB, Holcombe RG. 1987. Coefficient bias due to
specification search in econometric models. Atlantic
Economic Journal 15(3): 30–34.
Caudill SB, Holcombe RG. 1999. Specification search
and levels of significance in econometric models.
Eastern Economic Journal 25: 289–300.
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

Chalmers AF. 1999. What Is This Thing Called Science?
(3rd edn). Open University Press: Maidenhead, U.K.
Chamberlin TC. 1965 [1890]. The method of multiple
working hypotheses. Science 148(3671): 754–759.
Child J. 1972. Organizational structure, environment and
performance: the role of strategic choice. Sociology 6:
1–22.
Child J. 1997. Strategic choice in the analysis of action,
structure, organizations and environment: retrospect
and prospect. Organization Studies 18: 43–76.
Chiles TH, McMackin JF. 1996. Integrating variable risk
preferences, trust, and transaction cost economics.
Academy of Management Review 21: 73–99.
Collier A. 1994. Critical Realism: An Introduction to Roy
Bhaskar’s Philosophy. Verso: London, U.K.
Croson R, Anand J, Agarwal R. 2007. Using experiments
in corporate strategy research. European Management
Review 4: 173–181.
Cruickshank J (ed). 2003. Critical Realism: The Differ-
ence It Makes. Routledge: London, U.K.
Danermark B, Ekström M, Jakobsen L, Karlsson JC.
2002. Explaining Society: Critical Realism in the
Social Sciences. Routledge: London, U.K.
Davis GF, Marquis C. 2005. Prospects for organization
theory in the early twenty-first century: institutional
fields and mechanisms. Organization Science 16:
332–343.
De Cock C, Jeanes EL. 2006. Questioning consensus,
cultivating conflict. Journal of Management Inquiry
15: 18–30.
Denis DJ, Denis DK, Sarin A. 1999. Agency theory
and the influence of equity ownership structure
on corporate diversification strategies. Strategic
Management Journal 20(11): 1071–1076.
Donaldson L. 1995. American Anti-management Theories
of Organization: A Critique of Paradigm Proliferation.
Cambridge University Press: Cambridge, U.K.
Donaldson L. 1996. For Positivist Organization Theory:
Proving the Hard Core. Sage: London, U.K.
Douma S, George R, Kabir R. 2006. Foreign and
domestic ownership, business groups, and firm
performance: evidence from a large emerging market.
Strategic Management Journal 27(7): 637–657.
Downward P, Finch JH, Ramsay J. 2002. Critical
realism, empirical methods and inference: a critical
discussion. Cambridge Journal of Economics 26:
481–500.
Duhem P. 1954. The Aim and Structure of Physical
Theory. Princeton University Press: Princeton, NJ.
Dutton JE, Stumpf SA. 1991. Using behavioral simu-
lations to study strategic processes. Simulation and
Games 18: 207–224.
Easton G. 2000. Case research as a method for
industrial networks: a realist apologia. In Realist
Perspectives on Management and Organisations,
Ackroyd S, Fleetwood S (eds). Routledge: London,
U.K.; 205–219.
Fabian FH. 2000. Keeping the tension: pressures to
keep the controversy in the management discipline.
Academy of Management Review 25: 350–371.
Fairclough N. 2005. Discourse analysis in organization
studies. Organization Studies 26: 915–939.
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 155
Feige EL. 1975. The consequences of journal editorial
policies and a suggestion for revision. Journal of
Political Economy 83: 1291–1296.
Ferraro F, Pfeffer J, Sutton RI. 2005. Economics lan-
guage and assumptions: how theories can become self-
fulfilling. Academy of Management Review 30: 8–24.
Ferraro F, Pfeffer J, Sutton RI. 2009. How and why
theories matter: a comment on Felin and Foss (2009).
Organization Science 20: 669–675.
Fleetwood S. 2005. Ontology in organization and
management studies: a critical realist perspective.
Organization 12: 197–222.
Fleetwood S, Ackroyd S (eds). 2004. Critical Realist
Applications in Organisation and Management Stud-
ies. Routledge: London, U.K.
Gadamer H. 2002. Truth and Method (2nd revised edn).
Continuum: New York.
Ghoshal S. 2005. Bad management theories are destroy-
ing good management practices. Academy of Manage-
ment Learning and Education 4: 75–91.
Ghoshal S, Moran P. 1996. Bad for practice: a critique of
the transaction cost theory. Academy of Management
Review 21: 13–47.
Gist ME, Hopper H, Daniels D. 1998. Behavioral
simulation: application and potential in manage-
ment research. Organizational Research Methods 1:
251–295.
Godfrey PC, Hill CWL. 1995. The problem of unob-
servables in strategic management research. Strategic
Management Journal 16(7): 519–533.
Goldthorpe JH. 2001. Causation, statistics, and sociology.
European Sociological Review 17: 1–20.
Grant AM, Wall TD. 2009. The neglected science
and art of quasi-experimentation: why-to, when-to,
and how-to advice for organizational researchers.
Organizational Research Methods 12: 653–686.
Greenwald AG, Pratkanis AR, Leippe MR, Baumgard-
ner MH. 1986. Under what conditions does theory
obstruct research progress? Psychological Review 93:
216–229.
Grünbaum A. 1960. The Duhemian argument. Philosophy
of Science 27: 75–87.
Guala F, Mittone L. 2005. Experiments in economics:
external validity and the robustness of phenomena.
Journal of Economic Methodology 12: 495–515.
Hambrick DC. 2007. The field of management’s devotion
to theory: too much of a good thing? Academy of
Management Journal 50: 1346–1352.
Hamlin CL. 2002. Beyond Relativism: Raymond Boudon,
Cognitive Rationality and Critical Realism. Routlege:
London, U.K.
Harré R, Second PF. 1972. The Explanation of Social
Behavior. Basil Blackwell: Oxford, U.K.
Harrison D, Easton G. 2004. Temporally embedded
case comparison in industrial marketing research.
In Critical Realist Applications in Organisation and
Management Studies, Fleetwood S, Ackroyd S (eds).
Routledge: London, U.K.; 194–210.
Hedström P. 2005. Dissecting the Social: On the Prin-
ciples of Analytical Sociology. Cambridge University
Press: Cambridge, U.K.
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
156 K. D. Miller and E. W. K. Tsang
Hedström P, Swedberg R (eds). 1998. Social Mecha-
nisms: An Analytical Approach to Social Theory. Cam-
bridge University Press: Cambridge, U.K.
Hodgson GM. 1993. Economics and Evolution: Bringing
Life Back into Economics. University of Michigan
Press: Ann Arbor, MI.
Hofmann DA, Griffin MA, Gavin MB. 2000. The appli-
cation of hierarchical linear modeling to organizational
research. In Multilevel Theory, Research, and Methods
in Organizations: Foundations, Extensions, and New
Directions, Klein KJ, Kozlowski SWJ (eds). Jossey-
Bass: San Francisco, CA; 467–511.
Hough JR, White MA. 2003. Environmental dynamism
and strategic decision-making rationality: an exam-
ination at the decision level. Strategic Management
Journal 24(5): 481–489.
Hume D. 2000 [1739]. A Treatise of Human Nature.
Oxford University Press: Oxford, U.K.
Hunt SD. 1992. For reason and realism in marketing.
Journal of Marketing 56(2): 89–102.
Hutchison TW. 1988. The case for falsification. In
The Popperian Legacy in Economics, de Marchi N
(ed). Cambridge University Press: Cambridge, U.K.;
169–181.
Jackson N, Carter P. 1991. In defense of par-
adigm incommensurability. Organization Studies 12:
109–127.
Jarzabkowski P. 2008. Shaping strategy as a structuration
process. Academy of Management Journal 51:
621–650.
Johanson J, Vahlne J-E. 1977. The internationalization
process of the firm—a model of knowledge devel-
opment and increasing foreign market commitments.
Journal of International Business Studies 8: 23–32.
Kahneman D, Tversky A. 1979. Prospect theory: an
analysis of decision under risk. Econometrica 47:
263–291.
Kennedy PE. 2005. Oh no! I got the wrong sign! What
should I do? Journal of Economic Education 36:
77–92.
Klayman J, Ha YW. 1987. Confirmation, disconfirma-
tion, and information in hypothesis testing. Psycho-
logical Review 94: 211–228.
Knights D. 1992. Changing space: the disruptive impact
of a new epistemological location for the study of
management. Academy of Management Review 17:
514–536.
Kuhn TS. 1962. The Structure of Scientific Revolutions.
University of Chicago Press: Chicago, IL.
Kwan K-M, Tsang EWK. 2001. Realism and construc-
tivism in strategy research: a critical realist response
to Mir and Watson. Strategic Management Journal
22(12): 1163–1168.
Lakatos I. 1970. Falsification and the methodology of
scientific research programmes. In Criticism and the
Growth of Knowledge, Lakatos I, Musgrave A (eds).
Cambridge University Press: London, U.K.; 91–196.
Lane PJ, Cannella AA, Lubatkin MH. 1999. Ownership
structure and corporate strategy: one question viewed
from two different worlds. Strategic Management
Journal 20(11): 1077–1086.
Copyright  2010 John Wiley & Sons, Ltd.
La Porte TR, Consolini PM. 1991. Working in practice
but not in theory: theoretical challenges of ‘high relia-
bility organizations.’ Journal of Public Administration
Research and Theory 1: 19–47.
Law AM, Kelton WD. 2000. Simulation Modeling and
Analysis (3rd edn). McGraw-Hill: Boston, MA.
Lawson T. 1997. Economics and Reality. Routledge:
London, U.K.
Leamer E. 1978. Specification Searches: Ad Hoc
Inference with Nonexperimental Data. Wiley: New
York.
Lee AS. 1991. Integrating positivist and interpretive
approaches to organizational research. Organization
Science 2: 342–365.
Lewis MW, Grimes AJ. 1999. Metatriangulation: build-
ing theory from multiple paradigms. Academy of Man-
agement Review 24: 672–690.
Lieberson S. 1992. Einstein, Renoir, and Greeley: some
thoughts about evidence in sociology. American
Sociological Review 57: 1–15.
Lin Z, Hui C. 1999. Should lean replace mass
organization systems? A comparative examination
from a management coordination perspective. Journal
of International Business Studies 30: 45–80.
Lipton P. 1991. Inference to the Best Explanation.
Routledge: London, U.K.
Locke EA. 2007. The case for inductive theory building.
Journal of Management 33: 867–890.
Losee J. 2005. Theories on the Scrap Heap: Scientists
and Philosophers on the Falsification, Rejection, and
Replacement of Theories. University of Pittsburgh
Press: Pittsburgh, PA.
Machamer P, Darden L, Craver CF. 2000. Thinking
about mechanisms. Philosophy of Science 67: 1–25.
Mahoney JT. 1993. Strategic management and determin-
ism: sustaining the conversation. Journal of Manage-
ment Studies 30: 173–191.
Mäki U. 2000. Kinds of assumptions and their truth:
shaking an untwisted F-twist. Kyklos 53: 317–336.
Manicas PT. 1987. A History and Philosophy of the Social
Sciences. Basil Blackwell: Oxford, U.K.
Manicas PT. 2006. A Realist Philosophy of Social
Science: Explanation and Understanding. Cambridge
University Press: Cambridge, U.K.
Meehl PE. 1978. Theoretical risks and tabular asterisks:
Sir Karl, Sir Ronald, and the slow progress of
soft psychology. Journal of Consulting and Clinical
Psychology 46: 806–834.
Miller KD. 2005. The problem of method and the practice
of management research. In Research Methodology in
Strategy and Management (Vol. 2), Ketchen DJ Jr,
Bergh DD (eds). Elsevier Science: Amsterdam, The
Netherlands; 143–177.
Mingers J. 2001. Combining IS research methods:
towards a pluralist methodology. Information Systems
Research 12: 240–259.
Mingers J. 2004. Future directions in management
science modeling. In Critical Realist Applications in
Organisation and Management Studies, Fleetwood S,
Ackroyd S (eds). Routledge: London, U.K.; 164–193.
Mingers J. 2006. A critique of statistical modelling in
management science from a critical realist perspective:
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj

its role within multimethodology. Journal of the
Operational Research Society 57: 202–219.
Mir R, Watson A. 2000. Strategic management and the
philosophy of science: the case for a constructivist
methodology. Strategic Management Journal 21(9):
941–953.
Mir R, Watson A. 2001. Critical realism and construc-
tivism in strategy research: toward a synthesis. Strate-
gic Management Journal 22(12): 1169–1173.
Mitchell TR, James LR. 2001. Building better theory:
time and the specification of when things happen.
Academy of Management Review 26: 530–547.
Mohr LB. 1982. Explaining Organizational Behavior.
Jossey-Bass: San Francisco, CA.
Mone MA, McKinley W 1993. The uniqueness value and
its consequences for organization studies. Journal of
Management Inquiry 2: 284–296.
Nambisan S. 2002. Designing virtual customer environ-
ments for new product development: toward a theory.
Academy of Management Review 27: 392–413.
Nash R. 1999. What is real and what is realism in
sociology? Journal for the Theory of Social Behaviour
29: 445–466.
Nickel MN, Rodriguez MC. 2002. A review of research
on the negative accounting relationship between risk
and return: Bowman’s paradox. Omega 30: 1–18.
Nooteboom B. 1986. Plausibility in economics. Eco-
nomics and Philosophy 2: 197–224.
Numagami T. 1998. The infeasibility of invariant laws in
management studies: a reflective dialogue in defense
of case studies. Organization Science 9: 2–15.
Orlikowski WJ. 1992. The duality of technology:
rethinking the concept of technology in organizations.
Organization Science 3: 398–427.
Pajunen K. 2008. The nature of organizational mecha-
nisms. Organization Studies 29: 1449–1468.
Peacock MS. 2000. Explaining theory choice: an
assessment of the critical realist contribution to
explanation in science. Journal for the Theory of Social
Behaviour 30: 319–339.
Pentland BT. 1999. Building process theory with
narrative: from description to explanation. Academy
of Management Review 24: 711–724.
Perrow C. 1984. Normal Accidents: Living with High-risk
Technologies. Basic Books: New York.
Perrow C. 1994. The limits of safety: the enhancement
of a theory of accidents. Journal of Contingencies and
Crisis Management 2: 212–220.
Petersen B, Pedersen T. 1997. Twenty years after—sup-
port and critique of the Uppsala internationalization
model. In The Nature of the International Firm,
Björkman I, Forsgren M (eds). Copenhagen Business
School Press: Copenhagen, Denmark; 117–134.
Pfeffer J. 2007. A modest proposal: how we might change
the process and product of managerial research.
Academy of Management Journal 50: 1334–1345.
Polanyi M. 1962. Personal Knowledge: Toward a Post-
critical Philosophy. University of Chicago Press:
Chicago, IL.
Popper KR. 1959. The Logic of Scientific Discovery.
Basic Books: New York.
Copyright  2010 John Wiley & Sons, Ltd.
Testing Management Theories 157
Porter S. 1993. Critical realist ethnography: the case
of racism and professionalism in a medical setting.
Sociology 27: 591–609.
Powell TC. 2002. The philosophy of strategy. Strategic
Management Journal 23(9): 873–880.
Powell TC. 2003. Strategy without ontology. Strategic
Management Journal 24(3): 285–291.
Pozzebon M. 2004. The influence of a structurationist
view on strategic management research. Journal of
Management Studies 41: 247–272.
Putnam H. 1974. The ‘corroboration’ of theories. In The
Philosophy of Karl Popper, Schilpp PA (ed). Open
Court Publishing: LaSalle, IL; 221–240.
Quine WVO. 1961. From a Logical Point of View . Harper
and Row: New York.
Rahmandad H. 2008. Effect of delays on complexity
of organizational learning. Management Science 54:
1297–1312.
Rajagopalan N, Spreitzer GM. 1996. Toward a theory
of strategic change: a multi-lens perspective and
integrative framework. Academy of Management
Review 22: 48–79.
Rescher N. 1988. Rationality: A Philosophical Inquiry
into the Nature and the Rationale of Reason.
Clarendon Press: Oxford, U.K.
Rescher N. 2000. Realistic Pragmatism: An Introduction
to Pragmatic Philosophy. State University of New
York Press: Albany, NY.
Ricoeur P. 1981. Hermeneutics and the Human Sciences:
Essays on Language, Action and Interpretation.
Cambridge University Press: New York.
Roberts KH. 1989. New challenges in organization
research: high reliability organizations. Industrial
Crisis Quarterly 3: 111–125.
Robins JA. 1987. Organizational economics: notes on
the use of transaction-cost theory in the study of
organizations. Administrative Science Quarterly 32:
68–86.
Ron A. 2002. Regression analysis and the philosophy
of social science: a critical realist view. Journal of
Critical Realism 1: 119–142.
Sagan SD. 1993. The Limits of Safety: Organizations,
Accidents, and Nuclear Weapons. Princeton University
Press: Princeton, NJ.
Sarason Y, Dean T, Dillard JF. 2006. Entrepreneurship as
the nexus of individual and opportunity: a structuration
view. Journal of Business Venturing 21: 286–305.
Sarasvathy SD. 2001. Causation and effectuation: toward
a theoretical shift from economic inevitability to
entrepreneurial contingency. Academy of Management
Review 26: 243–263.
Sawyer KR, Beed C, Sankey H. 1997. Underdetermina-
tion in economics. The Duhem-Quine thesis. Eco-
nomics and Philosophy 13: 1–23.
Sayer A. 1992. Method in Social Science: A Realist
Approach (2nd edn). Routledge: London, U.K.
Sayer A. 2000. Realism and Social Science. Sage:
London, U.K.
Schwarz N, Groves RM, Schuman H. 1998. Survey
methods. In The Handbook of Social Psychology (4th
edn, Vol. 1), Gilbert DT, Fiske ST, Lindzey G (eds).
McGraw-Hill: Boston, MA; 143–179.
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj
158 K. D. Miller and E. W. K. Tsang
Schweiger DM, Sandberg WR, Rechner PL. 1989. Expe-
riential effects of dialectical inquiry, devil’s advocacy,
and consensus approaches to strategic decision mak-
ing. Academy of Management Journal 32: 745–772.
Searle JR. 2001. Rationality in Action. MIT Press:
Cambridge, MA.
Shadish WR, Cook TD, Campbell DT. 2002. Experimen-
tal and Quasi-experimental Designs for Generalized
Causal Inference. Houghton Mifflin: Boston, MA.
Søberg M. 2005. The Duhem-Quine thesis and experi-
mental economics: a reinterpretation. Journal of Eco-
nomic Methodology 12: 581–597.
Starbuck WH. 2006. The Production of Knowledge:
The Challenge of Social Science Research. Oxford
University Press: New York.
Stinchcombe AL. 1968. Constructing Social Theories.
Harcourt, Brace & World: New York.
Sutcliffe KM, Zaheer A. 1998. Uncertainty in the
transaction environment: an empirical test. Strategic
Management Journal 19(1): 1–23.
Szulanski G, Jensen RJ. 2004. Overcoming stickiness: an
empirical investigation of the role of the template in
the replication of organizational routines. Managerial
and Decision Economics 25: 347–363.
Tsang EWK. 2006. Behavioral assumptions and theory
development: the case of transaction cost economics.
Strategic Management Journal 27(11): 999–1011.
Tsang EWK, Kwan K-M. 1999. Replication and theory
development in organizational science: a critical realist
perspective. Academy of Management Review 24:
759–780.
Tsoukas H. 1989. The validity of idiographic research
explanations. Academy of Management Review 14:
551–561.
Copyright  2010 John Wiley & Sons, Ltd.
Van de Ven AH. 2007. Engaged Scholarship: A Guide
for Organizational and Social Research. Oxford
University Press: New York.
Wason PC, Johnson-Laird PN. 1972. Psychology of
Reasoning: Structure and Content. Harvard University
Press: Cambridge, MA.
Wathne KH, Heide JB. 2000. Opportunism in interfirm
relationships: forms, outcomes, and solutions. Journal
of Marketing 64(4): 36–51.
Webb EJ, Campbell DT, Schwartz RD, Sechrest L. 2000.
Unobtrusive Measures (revised edn). Sage: Thousand
Oaks, CA.
Weick KE. 1989. Theory construction as disciplined
imagination. Academy of Management Review 14:
516–531.
Wicks AC, Freeman RE. 1998. Organization studies
and the new pragmatism: positivism, anti-positivism,
and the search for ethics. Organization Science 9:
123–140.
Williamson OE. 1996. Economic organization: the case
for candor. Academy of Management Review 21:
48–57.
Wuisman JJJM. 2005. The logic of scientific discovery
in critical realist social scientific research. Journal of
Critical Realism 4: 366–394.
Yeung HW. 1997. Critical realism and realist research
in human geography: a method or a philosophy in
search of a method? Progress in Human Geography
21: 51–74.
Zhang Z, Zyphur MJ, Preacher KJ. 2009. Testing multi-
level mediation using hierarchical linear models: prob-
lems and solutions. Organizational Research Methods
12: 695–719.
Strat. Mgmt. J., 32: 139–158 (2010)
DOI: 10.1002/smj